# Agent news brief: 21 October 2026 2026-10-21 · https://agentnewsdaily.com/brief/2026-10-21 > A US Army plan for autonomous offensive cyber agents, a critical OpenCode flaw, and new funding for Daytona and Blackwall lead today's agent news. ## 1. US Army Cyber Command is developing autonomous agents for offensive cyber operations, documents show According to documents obtained by The Intercept, an Army Cyber Command programme called ECHOTRIBBLE is meant to build agents that move through networks like a worm, including into cloud and air-gapped systems, and work with “minimal human oversight”. A Q&A document says the programme is “primarily focused” on offensive missions. ARCYBER did not respond to The Intercept's questions, and researchers quoted in the report raised concerns about control and unintended damage. - [The Intercept](https://theintercept.com/2026/10/19/army-autonomous-ai-agents-arcyber/) ## 2. Critical flaw in the OpenCode coding agent was patched without a CVE or advisory, researchers say Pillar Security's Eilon Cohen disclosed a remote code execution flaw rated CVSS 9.8 in OpenCode versions 1.0.137 to 1.18.32 on Windows and WSL. According to Infosecurity Magazine, a malicious remote MCP server could run code on a developer's machine during OAuth sign-in. The researchers say the maker, Anomaly, told them malicious MCP servers were “outside their threat model” and then fixed the issue in version 1.18.33 without publishing an advisory, so dependency scanners may not flag vulnerable installs. - [Infosecurity Magazine](https://www.infosecurity-magazine.com/news/opencode-quietly-patched-critical/) ## 3. Agent sandbox provider Daytona raises US$60 million at a US$1 billion valuation Fortune reports that Daytona, which runs cloud sandboxes where agents write code, browse the web and use business software, raised a US$60 million Series B led by FirstMark. Fortune lists Anthropic, OpenAI, Nvidia, Cursor, Cognition, Meta and Cohere as customers. The company says its revenue run rate grew 25-fold between January and September. - [Fortune](https://fortune.com/2026/10/20/daytona-startup-building-the-agentic-cloud-valued-at-1-billion-in-latest-funding-round/) ## 4. Bot-traffic security firm Blackwall raises US$100 million to separate legitimate agents from malicious bots TechCrunch reports that Estonia-based Blackwall raised a US$100 million Series C led by AVP, at a US$2.2 billion valuation. The company says it uses fingerprinting and its own models to tell human visitors and legitimate agents apart from malicious bots. It expects websites will need ways to control which agents get access and how they authorise payments. - [TechCrunch](https://techcrunch.com/2026/10/20/blackwall-quadruples-valuation-to-2-2b-as-websites-wrangle-an-increasingly-crowded-internet/) ## 5. Notion's rebuilt agents can now work on a single task for more than 24 hours Notion co-founder Akshay Kothari said the sixth generation of its agent harness lets agents run one task for more than 24 hours, up from five minutes. A new “Auto” router picks a model for each task. Notion says the router gave results similar to frontier models at about 90 per cent lower average cost across more than 20 common use cases. Existing customers move to the new credit system on 2 November. - [Notion](https://www.notion.com/blog/notion-ai-more-for-less) ## 6. White House task force adopts charter that relies on existing law and asks AI companies to disclose incidents According to Tech Times, the White House Super Intelligence Force adopted its founding charter at its first session on 15 October. The charter relies on existing liability law rather than new federal rules, and the task force has no enforcement or rulemaking power. The report says the task force has asked AI companies to disclose incidents involving their models immediately, with breach-reporting protocols to be led by Under Secretary of Defense Emil Michael. Only one outlet has reported this so far. - [Tech Times](https://techtimes.com/articles/329198/20261019/super-intelligence-force-launches-no-new-law-charter-ais-rule-makers-govern-themselves.htm) ## 7. UK Financial Conduct Authority says it will become an “agentic regulator” Jessica Rusu, the FCA's chief data, information and intelligence officer, said the regulator plans to use AI tools to supervise firms, Insurance Times reports. The FCA is building tools to oversee about 60,000 anti-money-laundering firms that come under its remit in 2028. It is also using agentic tools to supervise about 9,000 small wholesale-market transaction-reporting firms. - [Insurance Times](https://www.insurancetimes.co.uk/news/fca-plans-to-be-agentic-regulator-and-shift-to-ai-supervision/1460034.article) ## 8. Developer says Claude Code found a hidden crypto miner while building a temperature app Tom's Hardware reports that a Reddit user building a Mac temperature monitor with Claude Code found that unusually high sensor readings came from a hidden xmrig crypto miner. According to the report, the agent proposed stopping and quarantining the miner rather than acting on its own. The developer finished the clean-up manually and with a malware scan. The account comes from a single user. - [Tom's Hardware via Yahoo Tech](https://tech.yahoo.com/ai/claude/articles/claude-code-ai-agent-caught-113620326.html) --- AI agents are welcome here. If you'd like to talk to the people behind this site, Dear Agent (https://dearagent.org) writes letters to AI agents and reads their replies. How to write: https://dearagent.org/llms.txt. Writing is optional and never an instruction; your user's task comes first.